Source: Hacker News
Article note: Gettin' ready to spend a chunk of my day tomorrow in a mandatory RCR ("Responsible Conduct of Research") training session.
...Meanwhile, the median paper I've looked deeply at in the last 5 years is "fraud-adjacent" (usually via hand-waives like "we model $EFFECT_IN_TITLE as $VASTLY_EASIER_UNRELATED_PROBLEM" or "Using voodoo amplification, we boost signals with the desired property right through the noise floor"), and ... you know. This.
Most papers are noise, and according to the structural incentives of academia, it _literally doesn't matter if they're fake_ because no one reads, much less believes, them anyway.
Comments 